A method for predicting and analyzing forest fires in different periods
By constructing a forest fire prediction model during the period of changes in forest fire prevention policies and combining multiple driver factors, the problem of ignoring policy impact in the existing technology is solved, and the accuracy of forest fire prediction is improved and management optimization is achieved.
Patent Information
- Application Number
- CN202311694022.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-11
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-12-11
AI Technical Summary
The existing technology ignores the impact of fire management policies on drivers in forest fire prediction, and it is difficult to analyze the changes in factors, resulting in poor prediction results.
Using the implementation time of forest fire prevention policy as the dividing line, obtain fire data from different periods, create sample point data through random points, combine factors such as terrain, vegetation, infrastructure, population, and meteorology, and use weak learning tree models to build a prediction model and analyze the model Changes in accuracy, fire probability and fire risk level.
The changes in model accuracy, fire prediction probability and fire risk level caused by policy changes have been deeply explored, providing a scientific basis for optimizing forest fire management and improving prediction accuracy.
Smart Images

Figure CN118095520B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of forest fire prediction and management, and in particular to a method for predicting and analyzing forest area fires in different periods. Background Art
[0002] Studying the driving factors of forest fires and their impacts is an important aspect of forest fire management and prevention, which can help fire prevention departments accurately assess forest fire hazards and effectively implement forest fire prevention strategies. Many studies have shown that the occurrence and spread of forest fires are affected by a combination of factors such as meteorology, vegetation (combustibles), terrain and human activities. Guo et al. found that the occurrence of forest fires is closely related to vegetation coverage, which has an important influence; Chang Yu et al. believed that terrain factors significantly change the distribution structure and quantity of vegetation, and to some extent will affect the occurrence and spread of forest fires; Di Liying et al. analyzed the spatial distribution characteristics of forest fires and found that roads, railways and residential areas are high-incidence areas of forest fires; Guo Futao et al. analyzed the driving factors of forest fires and found that population density and local GDP are significantly correlated with the occurrence of forest fires.
[0003] At present, some studies on forest fire prediction models have been carried out in the forests of northern China. Most of these studies focus on the comparison of forest fire driving factors between different climate gradients and between different geographical regions, and these studies will select fire point data within a certain period for modeling. At present, most of them simply consider factors such as meteorology, vegetation, topography, or different regions such as climate gradients and geographical regions. When modeling fires, they often ignore the impact of fire management policies on key information such as forest fire driving factors. Therefore, the analysis of changes in forest fire information from the perspective of fire management policy changes is still blank. Summary of the invention
[0004] 1. Technical issues to be resolved
[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for predicting and analyzing forest area fires in different periods, which solves the technical problem that the prior art solutions are difficult to undertake the analysis of the impact caused by changes in fire management policies and the changing laws of various factors.
[0006] (II) Technical solution
[0007] In order to achieve the above object, the main technical solutions adopted by the present invention include:
[0008] In a first aspect, an embodiment of the present invention provides a method for predicting and analyzing forest area fires at different periods, comprising:
[0009] Taking the implementation time of forest fire prevention policy as the dividing line, the fire data of forest areas in Inner Mongolia Autonomous Region in different periods before and after the dividing line were obtained;
[0010] Based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the demarcation line, random points are created and random date values are assigned at a ratio close to or greater than the forest fire points, and sample point data are formed based on the assigned random points and forest fire points.
[0011] Obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and combine with sample point data to obtain a sample data set;
[0012] The sample data set is recursively cut into multiple sub-data sets for modeling, and then a linear regression method is used to continuously construct weak learning tree models for the multiple sub-data sets, and the multiple continuous weak learning tree models are combined into a prediction model;
[0013] Based on the prediction model, fire data of different periods before and after the dividing line are predicted and analyzed to obtain at least one of the change information of model accuracy, fire prediction probability, fire driving factors and fire risk level under forest fire prevention policies in different periods.
[0014] Optionally, based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the demarcation line, random points are created with a ratio close to or greater than the forest fire points and random date assignment is performed, and sample point data is formed based on the obtained assigned random points and forest fire points, including:
[0015] Traverse the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the boundary line, and delete the fire data without geographical location;
[0016] Assign the remaining known fire points a value of 1;
[0017] Based on the national land use data, the distribution range of forest types in Inner Mongolia Autonomous Region was extracted, random points were created within the distribution range of forest types in Inner Mongolia Autonomous Region, and the random points were assigned a value of 0;
[0018] Random dates are assigned to the created random points to ensure that the generation of random points is completely random in time and space;
[0019] The minimum allowed distance between the created random points is set to avoid the created random points coinciding with or being adjacent to known fire points;
[0020] Sample point data are formed based on forest fire points and random points.
[0021] Optionally, obtaining driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source and combining the sample point data to obtain a sample data set includes:
[0022] Obtain GDEMV230M digital elevation data from the geospatial data cloud, and perform preprocessing on the GDEMV230M digital elevation data including spatial correction splicing, clipping, and resampling to obtain terrain factor data;
[0023] Obtain remote sensing monitoring data maps of China's land use status from the National Earth System Science Data Center, and extract forest type factor data corresponding to each fire point and random point from the remote sensing monitoring data maps of China's land use status;
[0024] Call the infrastructure data of the railways, roads and settlements closest to the fire point from the national geographic information resource directory service system;
[0025] Demographic and socioeconomic data were obtained from the Statistics Bureau of Inner Mongolia Autonomous Region;
[0026] The meteorological data of the spring fire prevention period and the autumn fire prevention period of the year before the fire was obtained from the meteorological service platform. At the same time, the average temperature, maximum temperature, average relative humidity, average surface temperature, maximum surface temperature, minimum relative humidity, average wind speed, sunshine hours and cumulative precipitation were selected on a daily scale. On a monthly scale, the monthly average temperature, monthly average relative humidity, monthly average surface temperature, monthly cumulative sunshine hours and monthly average precipitation of the month when the fire occurred and the month before the fire were selected;
[0027] The driving factor data including topography, forest type, infrastructure, population and socio-economics, and meteorology are combined with the corresponding sample point data to obtain the sample data set.
[0028] Optionally, predicting the fire data of different periods before and after the dividing line based on the prediction model includes:
[0029] The prediction model was built using the construction tools it relied on, and the prediction model was tested for discarding unimportant variables based on fire data in the forest area of Inner Mongolia Autonomous Region at different times. In the process of continuously adjusting the model layer parameters for the purpose of maximum accuracy, significant variables that were not discarded three times or more were selected to be retained.
[0030] With the help of the construction tools relied on by the enhanced regression tree model, the relative importance of the significant variables corresponding to the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line was obtained, and the significant variables were ranked according to their relative importance.
[0031] Significant variables are introduced into the fitting stage of the prediction model to obtain the prediction results of the prediction model in different periods before and after the dividing line.
[0032] Optionally, the prediction model is:
[0033]
[0034] Where, T m (X, γ m ) represents the mth weak learning tree model, m∈M, X represents the driving factor prediction variable, γ m represents the predictor variable parameter, which is the value assigned to the split point and leaf node of the decision tree. m The process is the learning and training process of a single weak learning tree model.
[0035] Optionally, fire data of different periods before and after the dividing line are analyzed based on the prediction model to obtain at least one of information on changes in model accuracy, fire prediction probability, fire driving factors, and fire risk level under forest fire prevention policies of different periods, including:
[0036] Compare the model's predictions with the observed values for the corresponding period to generate the model's residuals;
[0037] Several different critical values were set to calculate a series of sensitivity and specificity values, with sensitivity as an ordinal number and specificity as a negative number to obtain the receiver operating characteristic curve (ROC) under the prediction model;
[0038] Calculate the value of the area under the receiver operating characteristic curve ROC AUC;
[0039] The best critical value was obtained by calculating the Yoden index according to the sensitivity and specificity of the receiver operating characteristic curve ROC as the optimal threshold;
[0040] According to the residual error of the model, the area under the receiver operating characteristic curve (ROC) (AUC), and the optimal threshold, the prediction accuracy change information of the prediction model at different periods before and after the dividing line is determined;
[0041] The prediction results of the prediction model in different periods before and after the dividing line are analyzed to obtain the change information of fire prediction probability, fire driving factors and fire risk level.
[0042] Optionally,
[0043] The information on changes in fire driving factors under forest fire prevention policies in different periods includes: changes in significant variables of driving factors and changes in the ranking of significant variables;
[0044] Information on changes in fire risk levels under forest fire prevention policies in different periods includes: changes in the proportion of fire risk areas to forest areas.
[0045] In a second aspect, an embodiment of the present invention provides a forest area fire prediction and analysis system at different periods, including:
[0046] The fire data acquisition module is used to obtain the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the implementation time of the forest fire prevention policy as the dividing line;
[0047] The sample point generation module is used to create random points and perform random date assignment based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the demarcation line, and to form sample point data based on the obtained random points and forest fire points.
[0048] A sample data set generation module is used to obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and obtain a sample data set by combining sample points;
[0049] A prediction model building module is used to recursively cut the sample data set into multiple groups of sub-data sets for modeling, and then use the linear regression method to continuously build weak learning tree models for the multiple groups of sub-data sets, and combine the established multiple continuous weak learning tree models into a prediction model;
[0050] The change information output module is used to predict and analyze fire data in different periods before and after the dividing line based on the prediction model, and obtain at least one of the change information of model accuracy, fire prediction probability change information, fire driving factor change information and fire risk level change information under the forest fire prevention policy in different periods.
[0051] In a third aspect, an embodiment of the present invention provides a forest area fire prediction and analysis device at different periods, including:
[0052] at least one database;
[0053] and a memory in communication with the at least one database;
[0054] Wherein, the memory stores instructions that can be executed by the at least one database, and the instructions are executed by the at least one database so that the at least one database can execute the forest area fire prediction and analysis method at different periods as described above.
[0055] In a fourth aspect, an embodiment of the present invention provides a computer-readable medium having computer-executable instructions stored thereon, which, when executed by a processor, implement a forest area fire prediction and analysis method for different periods as described above.
[0056] (III) Beneficial effects
[0057] The beneficial effect of the present invention is: the present invention chooses to construct a prediction model suitable for the forest fire prediction scheme in Inner Mongolia Autonomous Region. Starting from the perspective of forest fire prevention policy, the present invention deeply explores the respective changing laws and internal connections of model accuracy, fire prediction probability, fire driving factors and fire risk level caused by policy changes, provides a reference for constructing fire prediction models of different time series, and provides a scientific basis for optimizing forest fire management policies and prediction and forecasting work. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] Figure 1 A schematic diagram of a flow chart of a method for predicting and analyzing forest area fires at different periods provided by an embodiment of the present invention;
[0059] Figure 2 A specific flow chart of step S2 of a method for predicting and analyzing forest area fires at different periods provided by an embodiment of the present invention;
[0060] Figure 3 A schematic diagram of a portion of step S5 of a method for predicting and analyzing forest area fires at different periods provided by an embodiment of the present invention;
[0061] Figure 4 A schematic diagram of the importance ranking of driving factors affecting the occurrence of forest fires according to a method for predicting and analyzing forest area fires at different periods provided by an embodiment of the present invention;
[0062] Figure 5 Another partial flow chart of step S5 of a method for predicting and analyzing forest area fires at different periods provided by an embodiment of the present invention;
[0063] Figure 6 A schematic diagram of the AUC values of a forest area fire prediction analysis method at different periods based on sample data sets at four different periods provided by an embodiment of the present invention;
[0064] Figure 7 A schematic diagram of the number of forest fires and burned areas from 1981 to 2020, providing a forest area fire prediction and analysis method for different periods in an embodiment of the present invention. DETAILED DESCRIPTION
[0065] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.
[0066] like Figure 1 As shown, a method for predicting and analyzing forest area fires at different periods proposed in an embodiment of the present invention includes: taking the implementation time of a forest fire prevention policy as a dividing line, obtaining fire data of forest areas in the Inner Mongolia Autonomous Region at different periods before and after the dividing line; creating random points at a ratio close to or greater than the forest fire points and assigning random dates based on the acquired fire data of forest areas in the Inner Mongolia Autonomous Region at different periods before and after the dividing line, and forming sample point data based on the assigned random points and forest fire points; obtaining driving factor data including terrain, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and obtaining a sample data set in combination with the sample point data; recursively cutting the sample data set into multiple groups of sub-data sets for modeling, and then continuously constructing weak learning tree models for the multiple groups of sub-data sets using a linear regression method, and combining the established multiple continuous weak learning tree models into a prediction model; predicting and analyzing the fire data at different periods before and after the dividing line based on the prediction model, and obtaining at least one of the change information of model accuracy, fire prediction probability change information, fire driving factor change information, and fire risk level change information under forest fire prevention policies at different periods.
[0067] The present invention chooses to construct a prediction model suitable for the forest fire prediction scheme in Inner Mongolia Autonomous Region. Starting from the perspective of forest fire prevention policy, it deeply explores the respective changing laws and internal connections of model accuracy, fire prediction probability, fire driving factors and fire risk level caused by policy changes, provides a reference for constructing fire prediction models of different time series, and provides a scientific basis for optimizing forest fire management policies and prediction and forecasting work.
[0068] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.
[0069] Specifically, the present invention provides a method for predicting and analyzing forest area fires at different periods, comprising:
[0070] S1. Taking the implementation time of forest fire prevention policy as the dividing line, obtain the fire data of forest areas in Inner Mongolia Autonomous Region at different periods before and after the dividing line.
[0071] The Inner Mongolia Autonomous Region is located in the northern border of my country (97°10′~126°29′E, 37°20′~53°20′N) Figure 1), is long and narrow, stretching from northeast to southwest, across three major regions: North China, Northeast China and Northwest China. The total land area of Inner Mongolia Autonomous Region reaches 1.183 million km2, equivalent to 12.3% of the national land area. It is a vast and historic region with complex and diverse terrain, unique climatic conditions, rich natural resources, low economic development level, obvious regional differences and ethnic distribution characteristics.
[0072] The landforms of the whole region are mainly plateau-type, and the altitude of most areas is above 1,000 meters. The west is mainly plains, and the east is mainly mountainous. The Inner Mongolia Autonomous Region is a region with unique geographical location and topographical features, and has obvious temperate continental monsoon climate characteristics. The temperature rises sharply in spring, and there are many strong winds; the summer is warm and the precipitation is concentrated; the temperature drops sharply in autumn, which often leads to the early formation of autumn frost; the winter is long and cold, with many cold waves. The annual precipitation is 350-450mm, mainly from June to August, the frost-free period is 80-150 days, and the average annual sunshine time exceeds 2,700 hours. The forest area in the whole region is 43.9493 million hm 2 , accounting for 37.15% of the autonomous region’s area; forest area is 24.879 million hm2 2 The forest area is 53.84% of the total forest area (including forest area and shrub area specially stipulated by the state). The mountain forests in the eastern part of Inner Mongolia Autonomous Region are mainly secondary forests and artificial forests. The main forest types are Larix gmelinii, Pinus sylvestris, Pinus tabuliformis, Betula platyphylla, Quercus mongolica, etc.
[0073] The embodiment of the present invention adopts the historical fire data of the Inner Mongolia Autonomous Region from 1981 to 2020 (40 years), and divides it into four periods (1981-March 14, 1988, March 15, 1988-2008, 2009-2020, 1981-2020) based on the implementation time of China's forest fire policy, and obtains data of the target area corresponding to the four periods.
[0074] S2. Based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the dividing line, random points are created and random dates are assigned at a ratio close to or greater than that of forest fire points, and sample point data are formed based on the assigned random points and forest fire points.
[0075] Furthermore, if Figure 2 As shown, step S1 includes:
[0076] S11. Traverse the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the boundary line, and delete the fire data without geographical location.
[0077] S12. Assign a value of 1 to the known fire points that are retained.
[0078] S13. Extract the distribution range of forest types in Inner Mongolia Autonomous Region based on national land use data, create random points within the distribution range of forest types in Inner Mongolia Autonomous Region, and assign the random points a value of 0.
[0079] S14. Random dates are assigned to the created random points to ensure that the generation of the random points is completely random in time and space.
[0080] S15. Set the minimum allowed distance between the created random points to avoid the created random points from coinciding with or being adjacent to known fire points.
[0081] S16. Form sample point data based on forest fire points and random points.
[0082] The forest fire data is provided by the Forestry and Grassland Bureau of the Inner Mongolia Autonomous Region, and contains detailed information on each forest fire, including geographic location, fire time, fire cause, casualties, etc. (Before 1990, the location of the fire was determined by the head of the forest farm, who identified each fire through a comprehensive method of fixed observation points in the forest combined with forest resource distribution maps. After 1990, the Global Positioning System (GPS) was used to record the location of the fire. A total of 3,631 fires were recorded from 1981 to 2020, of which 115 fires had no geographic coordinate information; this invention deletes fires without geographic location.
[0083] The prediction model used in the present invention requires the target variable to be binary, so the known fire point is assigned to 1, and then ArcGIS 10.8 is used to create random points in the study area and assigned to 0 (in order to avoid data being too discrete, random points are usually created with a ratio close to or more than the fire point. The present invention adopts fire point: random point = 1: 1.5). After creating the random points, random dates are assigned to non-fire points in Excel to ensure that the generation of random points is completely random in time and space. In order to avoid the created random points coinciding with or being adjacent to the known fire points, the minimum allowed distance between any two random points is set to 1 km. Establishing random points is to ensure that the position of the random points falls on the forest type. Therefore, based on the national land use data, the distribution range of forest types in the Inner Mongolia Autonomous Region is extracted, and random points are created within this range. Forest fire points and random points together constitute sample points.
[0084] S3. Obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and combine the driving factor data with the sample point data to obtain a sample data set, that is, bind the driving factor data to the corresponding sample point data to obtain the sample data set.
[0085] Further, step S3 includes:
[0086] The GDEMV230M digital elevation data was obtained from the geospatial data cloud, and the GDEMV230M digital elevation data was preprocessed including spatial correction splicing, clipping, and resampling to obtain terrain factor data.
[0087] The remote sensing monitoring data map of China's land use status was obtained from the National Earth System Science Data Center, and the forest type factor data corresponding to each fire point and random point were extracted from the remote sensing monitoring data map of China's land use status.
[0088] The infrastructure data including the railways, roads and settlements closest to the fire point are called from the national geographic information resource directory service system.
[0089] Demographic and socioeconomic data were obtained from the Statistics Bureau of Inner Mongolia Autonomous Region.
[0090] Meteorological data including the spring fire prevention period and autumn fire prevention period of the year before the fire were obtained from the meteorological service platform. At the same time, on the daily scale, the average temperature, maximum temperature, average relative humidity, average surface temperature, maximum surface temperature, minimum relative humidity, average wind speed, sunshine hours and cumulative precipitation were selected. On the monthly scale, the monthly average temperature, monthly average relative humidity, monthly average surface temperature, monthly cumulative sunshine hours and monthly average precipitation of the month when the fire occurred and the month before the fire were selected.
[0091] The driving factor data including topography, forest type, infrastructure, population and socio-economics, and meteorology are combined with the corresponding sample point data to obtain the sample data set.
[0092] In the specific embodiment, various types of driving factor data are introduced respectively:
[0093] (1) Terrain data: The DEM data comes from the GDEMV230M digital elevation data of the Geospatial Data Cloud (http: / / www.gscloud.cn). The DEM is processed by spatial correction, splicing, clipping, and resampling. The “Surface” tool in the Spatial Analyst toolkit in ArcGIS 10.8 software is used to extract the slope and aspect data of the study area. The aspect index is calculated according to the following formula:
[0094] Aspectindex=cos((0×2PI) / 360
[0095] In the formula, Aspect index is the slope index, θ is the slope value (0 to 360 degrees), and PI is the pi. The larger the slope index, the more sunny the slope. Finally, use the "Extract" tool to extract the elevation, slope, and slope aspect of the sample point.
[0096] (2) Forest type data: The forest type data comes from the National Earth System Science Data Center. The forest type of the study area is obtained by downloading the China Land Use Status Remote Sensing Monitoring Data Map with a resolution of 30m. ArcGIS10.8 is used to extract the forest type corresponding to each fire point and random point from the map according to the land use status map. The present invention uses the China Monthly NDVI dataset from 1981 to 2020 with a resolution of 5km. ArcGIS10.8 is also used to extract the NDVI values of the month corresponding to the fire points and random points. Among them, Normalized Difference Vegetation Index (NDVI): The occurrence of fire is closely related to the vegetation conditions at a specific time.
[0097] (3) Infrastructure data: The infrastructure data comes from the 1:250,000 national basic geographic database in the national geographic information resource directory service system. Specifically, the variables such as railways, roads, and settlements closest to the fire point were retrieved.
[0098] (4) Population and socioeconomic data: County-based population data, population density, GDP, and GDP per capita from 1985 to 2020 are provided by the Inner Mongolia Statistical Yearbook, and data from 1981 to 1985 are provided by the Inner Mongolia Autonomous Region Statistics Bureau. For details of the above variables, see Table 1.
[0099] (5) Meteorological data: Meteorological factors are generally considered to be the determining factors affecting the occurrence of forest fires. Studies have shown that since forest fires are often affected by the climate factors of the previous year's fire season, previous studies have mostly used the average value of meteorological elements in all fire prevention periods in a year as the dependent variable, without refining the meteorological elements of each fire prevention period in a year. In the present invention, the average temperature, average surface temperature, average humidity, and cumulative sunshine hours of the spring fire prevention period and autumn fire prevention period (there are two fire prevention periods in the Inner Mongolia Autonomous Region throughout the year before the fire occurred) are added.
[0100] In addition to the climatic factors during the fire prevention period, meteorological factors on the daily and monthly scales of the fire are also considered. On the daily scale, the average temperature (℃), maximum temperature (℃), average relative humidity (%), average surface temperature (℃), maximum surface temperature (℃), minimum relative humidity (%), average wind speed (m / s), sunshine hours, and cumulative precipitation (mm / 24h) are selected. On the monthly scale, the present invention selects the monthly average temperature (℃), monthly average relative humidity (%), monthly average surface temperature (℃), monthly cumulative sunshine hours, and monthly average precipitation (mm) in the month when the fire occurred and the month before the fire. See Table 1 below:
[0101] Table 1 Overview of driving factors
[0102]
[0103]
[0104] S4. Recursively cut the sample data set into multiple sub-data sets for modeling, then use the linear regression method to continuously construct weak learning tree models for the multiple sub-data sets, and combine the established multiple continuous weak learning tree models into a prediction model.
[0105] Further, step S4 includes:
[0106] Multiple consecutive weak learning tree models are combined in an additive manner to obtain a prediction model.
[0107] The model building method provided by the present invention combines the advantages of the regression tree algorithm and the boosting method, can adapt to complex nonlinear relationships, and automatically handle the interaction effects between factor variables, and can improve the stability and accuracy of the model. The regression tree algorithm recursively cuts the data set into multiple groups of sub-data sets that are easy to model, and then uses the linear regression method to construct weak learning trees for the sub-data sets; the boosting algorithm first constructs a prediction function series, and then combines each function sequence into a prediction function according to a certain rule to improve the accuracy of the weak classification algorithm.
[0108] The weak learning tree mentioned above refers to a tree model with a depth less than the set threshold. Each tree is only responsible for predicting part of the data. In this way, the model occupies less memory and has a faster prediction speed. Multiple tree models are then combined to build a model with strong computing power.
[0109] Preferably, the prediction model can be written as the sum of M classification and regression trees:
[0110]
[0111] Where, T m (X, γ m) represents the mth weak learning tree model, m∈M, X represents the driving factor prediction variable, γ m represents the predictor variable parameter, which is the value assigned to the split point and leaf node of the decision tree. m The process is the learning and training process of a single weak learning tree model.
[0112] S5. Predict and analyze the fire data of different periods before and after the dividing line based on the prediction model, and obtain at least one of the change information of the model accuracy, the change information of the fire prediction probability, the change information of the fire driving factor and the change information of the fire risk level under the forest fire prevention policy in different periods. Predict the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line based on the prediction model, and obtain the prediction results of each period.
[0113] Furthermore, if Figure 3 As shown, in step S5, predicting the fire data of different periods before and after the dividing line based on the prediction model includes:
[0114] S51. Using the construction tools that the prediction model relies on, the prediction model is tested for discarding unimportant variables based on fire data in the forest area of Inner Mongolia Autonomous Region at different periods. In the process of continuously adjusting the model layer parameters for the purpose of maximum accuracy, significant variables that have not been discarded three times or more are selected to be retained.
[0115] In Rstudio, the “gbm.simplify” function in the “gbm” package was used to test the abandonment of unimportant variables on the five training samples, and the parameter model accuracy was adjusted to achieve the highest accuracy and the variables that were not abandoned three times or less in the five subsets were selected and retained (Table 2). Finally, 16, 16, 16, and 16 variables were included in the model fitting stage as the main driving factors of forest fires in the four periods, respectively. The key parameters of the prediction model for the four periods are shown in Table 3.
[0116] Table 2 Selected significant variables in four periods through the prediction model
[0117]
[0118]
[0119] Note: (a) 1981-March 14, 1988 (b) March 15, 1988-2008 (c) 2009-2020 (d) 1981-2020.
[0120] Table 3 Key fitting parameters of the prediction model
[0121]
[0122] S52. With the help of the construction tools that the enhanced regression tree model relies on, the relative importance of the significant variables corresponding to the fire data in the forest area of Inner Mongolia Autonomous Region at different periods before and after the dividing line is obtained, and the significant variables are ranked according to their relative importance.
[0123] refer to Figure 4 , from the results of variable sorting of the whole sample, the daily temperature range and the daily minimum relative humidity are significant in the four periods. On a monthly scale, the average temperature of the month when the fire occurred, the precipitation of the month when the fire occurred, the average temperature of the month, and the precipitation of the month when the fire occurred are important variables affecting the occurrence of fires. Among the climate factors, the present invention also found that some indicators of the spring and autumn fire prevention periods have a significant impact on the occurrence of fires in the four periods, such as the cumulative precipitation of the previous spring prevention, the cumulative precipitation of the previous autumn prevention, and the highest surface temperature of the previous spring prevention; the altitude in the terrain factors shows a significant impact on the occurrence of fires in all models of the four periods; in terms of vegetation factors, NDVI is considered important in all four periods; forest type has a significant impact in the last three periods; in terms of socio-economic factors, the first period shows that per capita GDP and population density significantly affect fires; in terms of human factors: the four periods all show that the distance to the nearest residential area from the fire point, the distance to the nearest road from the fire point, and the distance to the nearest railway from the fire point have a significant impact.
[0124] S53. Introduce significant variables to participate in the fitting stage of the prediction model to obtain the prediction results of the prediction model in different periods before and after the dividing line.
[0125] Furthermore, if Figure 5 As shown, in step S5, the fire data of different periods before and after the dividing line are analyzed based on the prediction model, and at least one of the change information of the model accuracy, the change information of the fire prediction probability, the change information of the fire driving factor and the change information of the fire risk level under the forest fire prevention policy in different periods is obtained, including:
[0126] S54. Compare the prediction results of the model with the observed values of the corresponding period to generate the residual of the model.
[0127] In order to better evaluate the goodness of fit of the model, the predicted probability of the model was compared with the observed value, and the following results were obtained based on the generated residual analysis: in the first period, the enhanced regression tree model had clustered negative residual (over-prediction) areas in the northeast of Xilinguole League and the south of Chifeng City; in the second period, the enhanced regression tree model showed that there were negative residual areas in the northeast of Xilinguole League, and smaller positive residual (under-prediction) areas appeared in the rest of the areas; in the third period, the model showed that there were partial negative residuals in the southeast and south of Hulunbuir and the south of Xilinguole League, and smaller positive residual areas appeared in the rest of the areas; in all years, the model showed that there were positive residual areas in the southwest of Hulunbuir City, the central and southern parts of Xilinguole League.
[0128] S55. Set several different critical values to calculate a series of sensitivity and specificity values, take sensitivity as an ordinal number, and take specificity as a negative number to obtain the receiver operating characteristic curve (ROC) under the prediction model.
[0129] S56. Calculate the value of the area under the receiver operating characteristic curve (ROC), AUC.
[0130] S57. The optimal critical value is calculated by calculating the Yoden index based on the sensitivity and specificity of the receiver operating characteristic curve (ROC) to serve as the optimal threshold.
[0131] S58. Determine the prediction accuracy change information of the prediction model in different periods before and after the dividing line based on the residual of the model, the area under the receiver operating characteristic curve (ROC) (AUC), and the optimal threshold.
[0132] The present invention uses ROC (receiver operating characteristic) and AUC (area undercurve) to test the prediction effect. AUC can be used as a numerical value to intuitively evaluate the quality of the classifier, and the larger the value, the better the classification effect. The classification (cut-off point) can be calculated by the Youden index formula (Youden index = sensitivity value + specificity value - 1), and then the prediction probability can be classified. If the calculated prediction probability value is greater than the threshold value, it is considered that a forest fire will occur, and if it is less than the threshold value, it is considered that a forest fire will not occur.
[0133] The present invention calculated the prediction accuracy of the data set at each period, the AUC value under the ROC curve and the corresponding cut-off value (Table 4) and drew the ROC curve diagrams at four different times ( Figure 6), to test the prediction ability of the prediction model in different periods. The results show that the AUC values of the prediction model in 4 different periods are all greater than 0.9406, and the prediction accuracies are all greater than 87.3%, indicating that the model fits well in 4 different periods. Generally speaking, the AUC values are all Period 1 < Period 2 < Period 3, and the AUC value in Period 3 is slightly higher than that of all years.
[0134] Table 4 Feature sets in the fitting of 5 random sample data and full sample data
[0135]
[0136] S59. Analyze the prediction results of the prediction model in different periods before and after the demarcation line to obtain information on the change of fire prediction probability, information on the change of fire driving factors, and information on the change of fire risk levels. Among them, the information on the change of fire driving factors under the forest fire prevention policies in different periods includes: the change of significant variables of driving factors and the change of the ranking of significant variables; the information on the change of fire risk levels under the forest fire prevention policies in different periods includes: the change of the proportion of fire risk areas in forest areas.
[0137] The present invention uses the prediction model to draw the fire occurrence probability maps of 4 periods in the Arcgis environment using Kriging ((a) March 14, 1981 - 1988 (Period 1); (b) March 15, 1988 - 2008 (Period 2); (c) 2009 - 2020 (Period 3); (d) 1981 - 2020). The probability varies between 0 and 1. The larger the value of the grid, the higher the probability of a fire occurring. Based on the fire occurrence probability map and the Youden index, draw the fire risk zoning maps of different periods, and divide them into five levels, namely: extremely low fire risk (0 < P < 0.2), low fire risk (0.2 < P < 0.4), medium fire risk (0.4 < P < 0.6), high risk (0.6 < P < 0.8), and extremely high risk (0.8 < P < 1).
[0138] The present invention has counted the total number of forest fires and the burned area that occurred in Inner Mongolia Autonomous Region from 1981 to 2020 ( Figure 7 ), and classified them into 3 categories according to the causes of forest fires: human factors (mainly including smoking, burning paper, production and living fires, etc.), natural factors (lightning fires), and external intrusion factors (including foreign burning and burning in other provinces and cities).
[0139] Figure 7It shows that there are obvious differences in the number of forest fires caused by different disaster factors in Inner Mongolia Autonomous Region from 1981 to 2020. Natural factors, external invasion factors, and human factors account for 25.6%, 2.8%, and 71.6% of the total number of forest fires in Inner Mongolia Autonomous Region, respectively. Among them, the number of forest fires caused by human factors, which accounted for the highest proportion, was 2,054, and the number of forest fires caused by external invasion factors, which accounted for the lowest proportion, was 79. In the four different periods, the main disaster factor for forest fires in Inner Mongolia Autonomous Region was human factors. Figure 7 As can be seen from Table 5, before March 15, 1988, the number of man-made fires was high and the fire area was large; however, after the formulation and revision of the "Regulations", the situation improved. The number of lightning fires increased significantly, from 6.9 per year from March 14, 1981 to March 14, 1988 to 20.19 from March 15, 1988 to 2008 and 18.13 from 2009 to 2020.
[0140] Table 5 Average annual number of fires and burned areas in different periods
[0141]
[0142]
[0143] Thus, the impact of fire management policies in different periods is mainly caused by human factors. With the formulation and improvement of the Regulations, the number of fires caused by human factors has decreased significantly, and the number of natural fires caused by lightning strikes has increased, which is one of the reasons for the change in fire risk areas. According to the analysis of forest fire risk levels in Inner Mongolia Autonomous Region, the probability of forest fires in the first and third periods is very different from the fire risk level. The main areas with a higher probability of forest fires in the first period are the northeast and southeast of Hulunbuir City, while the higher probability of fires in the third period in Hulunbuir City has shifted to the southern part of Hulunbuir, and the fire risk level has decreased and the area has shrunk. The fire risk levels in Xilin Gol League, Chifeng City, Tongliao City, Hohhot City, Ulanqab City and Ordos City have increased. The present invention finds that the distribution of forest fire risks is constantly expanding, mainly because in recent years, with the implementation of the national ecological construction project, vegetation has been restored, surface combustibles have accumulated rapidly, and forest areas with high fire risk levels have continued to expand.
[0144] The risk levels of forest fires in different regions of the Inner Mongolia Autonomous Region vary greatly. In order to achieve scientific and reasonable planning of forest fire prevention work, the present invention needs to carry out fire prevention planning according to local conditions, and then carry out scientific and reasonable planning of forest fire prevention work on this basis. First, in the fire prevention planning, it is necessary to strengthen the monitoring of high and high fire risk areas and the management of forest combustibles to fundamentally eliminate the hidden dangers of forest fires; secondly, it is necessary to effectively manage forest fires, effectively prevent and control forest fires, so as to reduce the hazards of forest fires; thirdly, it is necessary to increase the popularization of fire protection knowledge to the local people, establish the concept of sustainable development, implement hierarchical management, save manpower, material resources and financial resources, and at the same time improve the efficiency of forest fire fighting.
[0145] In addition, an embodiment of the present invention provides a forest area fire prediction and analysis system at different periods, including:
[0146] The fire data acquisition module is used to obtain the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the implementation time of the forest fire prevention policy as the dividing line.
[0147] The sample point generation module is used to create random points and assign random dates based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the dividing line, with a ratio close to or greater than the forest fire points, and form sample point data based on the assigned random points and forest fire points.
[0148] The sample data set generation module is used to obtain driving factor data including terrain, forest type, infrastructure, population and socio-economics, and meteorology from at least one data source, and obtain a sample data set in combination with sample points.
[0149] The prediction model building module is used to recursively cut the sample data set into multiple groups of sub-data sets for modeling, and then use the linear regression method to continuously build weak learning tree models for the multiple groups of sub-data sets, and combine the established multiple continuous weak learning tree models into a prediction model.
[0150] The change information output module is used to predict and analyze fire data in different periods before and after the dividing line based on the prediction model, and obtain at least one of the change information of model accuracy, fire prediction probability change information, fire driving factor change information and fire risk level change information under the forest fire prevention policy in different periods.
[0151] Furthermore, an embodiment of the present invention provides a device for predicting and analyzing forest area fires at different periods, comprising: at least one database; and a memory communicatively connected to the at least one database; wherein the memory stores instructions executable by the at least one database, and the instructions are executed by the at least one database so that the at least one database can execute a method for predicting and analyzing forest area fires at different periods as described above.
[0152] Furthermore, an embodiment of the present invention provides a computer-readable medium having computer-executable instructions stored thereon, and when the executable instructions are executed by a processor, the above-mentioned method for predicting and analyzing forest area fires at different periods is implemented.
[0153] In summary, the embodiments of the present invention provide a method, system, device and medium for predicting and analyzing forest area fires in different periods. The present invention uses China’s old and new forest fire policies as the dividing line, divides the historical forest fire data of the Inner Mongolia Autonomous Region into four stages, establishes a forest fire occurrence model for the Inner Mongolia Autonomous Region, predicts the occurrence of forest fires, and explains the differences in forest fires and fire risk changes in different periods. The prediction results show that (1) the modeling accuracy AUC of the four periods is greater than 0.94, showing excellent performance. The prediction model has a good ability to predict the occurrence of forest fires in Inner Mongolia Autonomous Region; (2) Indicators such as daily temperature difference, daily minimum relative humidity, cumulative precipitation in the previous spring, cumulative precipitation in the previous autumn and maximum surface temperature in the previous spring are identified as important driving factors affecting the occurrence of forest fires in Inner Mongolia Autonomous Region, and are always the dominant factors affecting forest fires in different periods. In this invention, factors such as altitude and the nearest road distance to the fire point are also important driving factors affecting the occurrence of forests and fires in Inner Mongolia Autonomous Region; (3) Before and after the implementation of the new and old "Forest Fire Prevention Regulations", the distribution of forest fire risks in Inner Mongolia Autonomous Region in four different periods was quite different. From 1981 to March 14, 1988, the medium, high and extremely high forest fire risk areas were distributed in the eastern part of Hulunbuir. From 2009 to 2020, the medium, high and extremely high forest fire risk areas are distributed in the southern and central parts of Hulunbuir, the southwestern part of Chifeng City, the central part of Xilin Gol League and Hohhot City, the southern part of Ulanqab City and Baotou City, and the eastern part of Ordos City. The present invention helps to understand the changes in the driving factors and fire risk levels of forest fires in the Inner Mongolia Autonomous Region under the influence of the "Forest Fire Prevention Regulations" in different periods, and provides a scientific basis for optimizing forest fire management policies and prediction and forecasting work.
[0154] Since the system / device described in the above embodiments of the present invention is a system / device used to implement the method of the above embodiments of the present invention, a person skilled in the art can understand the specific structure and deformation of the system / device based on the method described in the above embodiments of the present invention, and thus will not be described in detail here. All systems / devices used in the method of the above embodiments of the present invention belong to the scope of protection of the present invention.
[0155] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0156] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions.
[0157] It should be noted that in the claims, any reference numerals placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In the claims enumerating several means, several of these means may be embodied by the same hardware. The use of the words first, second, third, etc., is for convenience of expression only and does not indicate any order. These words may be understood as part of the component name.
[0158] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.
[0159] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments after knowing the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.
[0160] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention should also include these modifications and variations.
Claims
1. A method for predicting and analyzing forest area fires at different periods, characterized in that: include: Taking the implementation time of forest fire prevention policy as the dividing line, the fire data of forest areas in Inner Mongolia Autonomous Region in different periods before and after the dividing line were obtained; Based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the demarcation line, random points are created and random date values are assigned at a ratio close to or greater than the forest fire points, and sample point data are formed based on the assigned random points and forest fire points. Obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and combine with sample point data to obtain a sample data set; The sample data set is recursively cut into multiple sub-data sets for modeling, and then a linear regression method is used to continuously construct weak learning tree models for the multiple sub-data sets, and the multiple continuous weak learning tree models are combined into a prediction model; Based on the prediction model, fire data of different periods before and after the dividing line are predicted and analyzed to obtain at least one of the following information: change information of model accuracy, change information of fire prediction probability, change information of fire driving factors, and change information of fire risk level under forest fire prevention policies of different periods; Among them, according to the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line, random points are created with a ratio close to or greater than the forest fire points and random date assignment is performed. The sample point data formed according to the obtained assigned random points and forest fire points include: Traverse the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the boundary line, and delete the fire data without geographical location; Assign the remaining known fire points a value of 1; Based on the national land use data, the distribution range of forest types in Inner Mongolia Autonomous Region was extracted, random points were created within the distribution range of forest types in Inner Mongolia Autonomous Region, and the random points were assigned a value of 0; Random dates are assigned to the created random points to ensure that the generation of random points is completely random in time and space; The minimum allowed distance between the created random points is set to avoid the created random points coinciding with or being adjacent to known fire points; Form sample point data based on forest fire points and random points; Prediction of fire data at different periods before and after the dividing line based on the prediction model includes: The prediction model was built using the construction tools it relied on, and the prediction model was tested for discarding unimportant variables based on fire data in the forest area of Inner Mongolia Autonomous Region at different times. In the process of continuously adjusting the model layer parameters for the purpose of maximum accuracy, significant variables that were not discarded three times or more were selected to be retained. With the help of the construction tools relied on by the enhanced regression tree model, the relative importance of the significant variables corresponding to the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line was obtained, and the significant variables were ranked according to their relative importance. Introduce significant variables to participate in the fitting stage of the prediction model to obtain the prediction results of the prediction model in different periods before and after the dividing line; The prediction model is: Where, T m (X, γ m ) represents the mth weak learning tree model, m∈M, X represents the driving factor prediction variable, γ m represents the predictor variable parameter, which is the value assigned to the split point and leaf node of the decision tree. m The process is the learning and training process of a single weak learning tree model.
2. A method for predicting and analyzing forest area fires at different periods as claimed in claim 1, characterized in that: Obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and combine with sample point data to obtain a sample data set including: Obtain GDEMV230M digital elevation data from the geospatial data cloud, and perform preprocessing on the GDEMV230M digital elevation data including spatial correction splicing, clipping, and resampling to obtain terrain factor data; Obtain remote sensing monitoring data maps of China's land use status from the National Earth System Science Data Center, and extract forest type factor data corresponding to each fire point and random point from the remote sensing monitoring data maps of China's land use status; Call the infrastructure data of the railways, roads and settlements closest to the fire point from the national geographic information resource directory service system; Demographic and socioeconomic data were obtained from the Statistics Bureau of Inner Mongolia Autonomous Region; The meteorological data of the spring fire prevention period and the autumn fire prevention period of the year before the fire was obtained from the meteorological service platform. At the same time, the average temperature, maximum temperature, average relative humidity, average surface temperature, maximum surface temperature, minimum relative humidity, average wind speed, sunshine hours and cumulative precipitation were selected on a daily scale. On a monthly scale, the monthly average temperature, monthly average relative humidity, monthly average surface temperature, monthly cumulative sunshine hours and monthly average precipitation of the month when the fire occurred and the month before the fire were selected; The driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology are combined with the corresponding sample point data to obtain the sample data set.
3. A method for predicting and analyzing forest area fires at different periods as claimed in claim 1, characterized in that: Based on the prediction model, the fire data of different periods before and after the dividing line are analyzed to obtain at least one of the following information: information on the change of model accuracy, fire prediction probability, fire driving factors, and fire risk level under forest fire prevention policies in different periods: Compare the model's predictions with the observed values for the corresponding period to generate the model's residuals; Several different critical values were set to calculate a series of sensitivity and specificity values, with sensitivity as an ordinal number and specificity as a negative number to obtain the receiver operating characteristic curve (ROC) under the prediction model; Calculate the value of the area under the receiver operating characteristic curve ROC AUC; The best critical value was obtained by calculating the Yoden index according to the sensitivity and specificity of the receiver operating characteristic curve ROC as the optimal threshold; According to the residual error of the model, the area under the receiver operating characteristic curve (ROC) (AUC), and the optimal threshold, the prediction accuracy change information of the prediction model in different periods before and after the dividing line is determined; The prediction results of the prediction model in different periods before and after the dividing line are analyzed to obtain the change information of fire prediction probability, fire driving factors and fire risk level.
4. A method for predicting and analyzing forest area fires at different periods as claimed in claim 3, characterized in that: The information on changes in fire driving factors under forest fire prevention policies in different periods includes: changes in significant variables of driving factors and changes in the ranking of significant variables; Information on changes in fire risk levels under forest fire prevention policies in different periods includes: changes in the proportion of fire risk areas to forest areas.
5. A forest area fire prediction and analysis system for different periods, characterized in that: include: The fire data acquisition module is used to obtain the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the implementation time of the forest fire prevention policy as the dividing line; The sample point generation module is used to create random points and perform random date assignment based on the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the demarcation line, and to form sample point data based on the obtained random points and forest fire points. A sample data set generation module is used to obtain driving factor data including topography, forest type, infrastructure, population, socio-economics, and meteorology from at least one data source, and obtain a sample data set by combining sample points; A prediction model building module is used to recursively cut the sample data set into multiple groups of sub-data sets for modeling, and then use the linear regression method to continuously build weak learning tree models for the multiple groups of sub-data sets, and combine the established multiple continuous weak learning tree models into a prediction model; A change information output module is used to predict and analyze the fire data of different periods before and after the dividing line based on the prediction model, and obtain at least one of the change information of the model accuracy, the change information of the fire prediction probability, the change information of the fire driving factor and the change information of the fire risk level under the forest fire prevention policy of different periods; Among them, according to the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line, random points are created with a ratio close to or greater than the forest fire points and random date assignment is performed. The sample point data formed according to the obtained assigned random points and forest fire points include: Traverse the fire data of the forest area of Inner Mongolia Autonomous Region at different periods before and after the boundary line, and delete the fire data without geographical location; Assign the remaining known fire points a value of 1; Based on the national land use data, the distribution range of forest types in Inner Mongolia Autonomous Region was extracted, random points were created within the distribution range of forest types in Inner Mongolia Autonomous Region, and the random points were assigned a value of 0; Random dates are assigned to the created random points to ensure that the generation of random points is completely random in time and space; The minimum allowed distance between the created random points is set to avoid the created random points coinciding with or being adjacent to known fire points; Form sample point data based on forest fire points and random points; Prediction of fire data at different periods before and after the dividing line based on the prediction model includes: The prediction model was built using the construction tools it relied on, and the prediction model was tested for discarding unimportant variables based on fire data in the forest area of Inner Mongolia Autonomous Region at different times. In the process of continuously adjusting the model layer parameters for the purpose of maximum accuracy, significant variables that were not discarded three times or more were selected to be retained. With the help of the construction tools relied on by the enhanced regression tree model, the relative importance of the significant variables corresponding to the fire data of the forest area of Inner Mongolia Autonomous Region in different periods before and after the dividing line was obtained, and the significant variables were ranked according to their relative importance. Introduce significant variables to participate in the fitting stage of the prediction model to obtain the prediction results of the prediction model in different periods before and after the dividing line; The prediction model is: Where, T m (X, γ m ) represents the mth weak learning tree model, m∈M, X represents the driving factor prediction variable, γ m represents the predictor variable parameter, which is the value assigned to the split point and leaf node of the decision tree. m The process is the learning and training process of a single weak learning tree model.
6. A device for predicting and analyzing forest fires at different times, characterized in that: include: at least one database; and a memory in communication with the at least one database; Wherein, the memory stores instructions that can be executed by the at least one database, and the instructions are executed by the at least one database so that the at least one database can execute a forest area fire prediction and analysis method for different periods as described in any one of claims 1-4.
7. A computer-readable medium having computer-executable instructions stored thereon, characterized in that: When the executable instructions are executed by the processor, a forest area fire prediction and analysis method for different periods as described in any one of claims 1 to 4 is implemented.
